Variability and Predictability in Long- Range Predictions Arun Kumar Climate Prediction Center College Park, Maryland, USA arun.kumar@noaa.gov ICTP-IITM ESM Workshop 18 July, 2016 1/33
Outline • What is weather and climate variability? • What is predictability? • How is predictability quantified? • Sources of predictability • Estimating predictability • Realizing predictability (or prediction skill) ICTP-IITM ESM Workshop 18 July, 2016 2/33
Weather and Climate Variability • Temperature tomorrow is not the same as today • Monthly (seasonal) mean precipitation for June-July-August seasonal average over India is not the same in 2014 as in 2015 • Average precipitation over India for a 10-year average changes from one decade to another ICTP-IITM ESM Workshop 18 July, 2016 3/33
2014 2015 ICTP-IITM ESM Workshop 18 July, 2016 4/33
Annual Mean All India Temperature Anomaly ICTP-IITM ESM Workshop 18 July, 2016 5/33
Quantifying Variability ICTP-IITM ESM Workshop 18 July, 2016 6/33
Variance of 200-mb DJF Seasonal Mean Height ICTP-IITM ESM Workshop 18 July, 2016 7/33
Predictability • Predictability: From the knowledge of the current state of the ocean, our ability to anticipate its future evolution • Prediction for a particular time-scale, what fraction of variability can be anticipated? – Predictability varies between 0-100% of variability ICTP-IITM ESM Workshop 18 July, 2016 8/33
Why all the variability is not predictable? ICTP-IITM ESM Workshop 18 July, 2016 9/33
There is always a spread (uncertainty) in forecasts! • Non-linear dynamical systems sensitivity to specification of initial conditions • Deterministic chaos • Uncertainty could be better quantified, but can never be removed ICTP-IITM ESM Workshop 18 July, 2016 10/33
Example of Seasonal Prediction ICTP-IITM ESM Workshop 18 July, 2016 11/33
Example of Climate Projection ICTP-IITM ESM Workshop 18 July, 2016 12/33
• The forecast spread (uncertainty) can be quantified using ensemble prediction approach where a collection of forecasts is initiated from small perturbations in the initial conditions • In a nutshell – The reason for a limit on predictability stems from limits on the accuracy of predictions on shorter time-scales – One cannot always predict the state of the atmosphere ∆t from now with 100% accuracy no matter how small ∆t is. ICTP-IITM ESM Workshop 18 July, 2016 13/33
How is Predictability Quantified? • Spread in forecast outcomes from different initial conditions can be quantified as probability density function (PDF) • It is our ability to distinguish PDF of outcomes for the event to be predicted from the climatological PDF • Differences in the PDF can come from differences in various moments of the PDF – Mean – Spread – Skewness ICTP-IITM ESM Workshop 18 July, 2016 14/33
How is Predictability Quantified? Climatological PDF PDF for a Season (Red) ICTP-IITM ESM Workshop 18 July, 2016 15/33
High predictability ICTP-IITM ESM Workshop 18 July, 2016 16/33
Low predictability ICTP-IITM ESM Workshop 18 July, 2016 17/33
Why it is Important to Understand and Quantify Predictability? • Helps gauge limits of prediction skill and manage expectations • Helps pinpoint sources of predictability, e.g., SST for atmospheric variability • How do climate models simulate processes, physics and interactions to better predict “sources” of predictability? • Provides one way to focus model improvements • Where to place limited resources (ensemble size, model resolution, analysis, perturbations,…) ICTP-IITM ESM Workshop 18 July, 2016 18/33
Sources of Predictability ICTP-IITM ESM Workshop 18 July, 2016 19/33
Sources of Predictability Weather – Atmospheric initial conditions • Seasonal – Boundary conditions (upper oceans, soil moisture, snow, sea- ice…) • Decadal – deeper oceans,… • Climate projections – CO 2 ,… • For different lead time, the relative contribution from sources of predictability • differs ICTP-IITM ESM Workshop 18 July, 2016 20/33
Influence of Various Factors on the PDF … initial conditions … boundary conditions … external conditions ICTP-IITM ESM Workshop 18 July, 2016 21/33
Seasonal-to-Interannual - ENSO Sea Surface Temperature Anomaly ICTP-IITM ESM Workshop 18 July, 2016 22/33
Decadal - PDO Sea Surface Temperature Anomaly (shading) ICTP-IITM ESM Workshop 18 July, 2016 23/33
Estimating Predictability ICTP-IITM ESM Workshop 18 July, 2016 24/33
Methods for Estimating Predictability • Observational data Daily time-series – Predictor – Predictand relationships – Analogs – Daily time-series DJF Z700 Correlation with SST index • Simple; unbiased, but non-linearity is hard to incorporate ICTP-IITM ESM Workshop 18 July, 2016 25/33
Methods for estimating predictability • Models – Ensemble of integrations • Spread among the ensemble members is the unpredictable component • Ensemble mean (the common part) is the predictable component ICTP-IITM ESM Workshop 18 July, 2016 26/33
Model Simulations ICTP-IITM ESM Workshop 18 July, 2016 27/33
Decomposing Total Variability ICTP-IITM ESM Workshop 18 July, 2016 28/33
Ratio of Predictable and Unpredictable Component 200mb Z ICTP-IITM ESM Workshop 18 July, 2016 29/33
Realizing Predictability Predictability Prediction skill • Requires a real-time forecast system • To realize predictability that exists, forecast systems need to have certain • attributes Design and framework of long-range prediction systems (Thursday) • ICTP-IITM ESM Workshop 18 July, 2016 30/33
Realizing Predictability ICTP-IITM ESM Workshop 18 July, 2016 31/33
Implication of Limited Predictability • Since future outcomes are not certain, forecasts have to be probabilistic • Decision making under probabilistic information context is hard ICTP-IITM ESM Workshop 18 July, 2016 32/33
Summary • There is variability associated with all time-scales • All variability cannot be anticipated in advance – Predictability • There are physical reasons that allow us to anticipate variability – sources of predictability • Predictability can be estimated either from observational data or model simulations • Forecast systems allow to realize predictability as prediction skill ICTP-IITM ESM Workshop 18 July, 2016 33/33
Recommend
More recommend